Submucosal Tunneling Endoscopic Resection for Upper Gastrointestinal Subepithelial Lesions: Development of a Digital...

Background and Aims Submucosal tunneling endoscopic resection (STER) is a minimally invasive technique for treating subepithelial lesions (SELs) originating from the muscularis propria. We aimed to evaluate the efficacy and safety of STER, independent predictors of procedural outcomes, and establish evidence-based patient selection criteria. Methodology A retrospective analysis was conducted on patients who underwent STER for esophageal and gastric SELs between January 2017 and December 2024. En bloc resection was defined as complete dissection of the lesion as a single piece from its site of origin within the submucosal tunnel. Piecemeal retrieval referred to fragmentation of an en bloc-dissected specimen during transluminal extraction owing to size constraints. Primary outcomes included technical success, en bloc resection rates, and complication rates. Secondary outcomes assessed factors predictive of piecemeal retrieval, procedure difficulty, and complication risk. Multivariable logistic regression and receiver-operating characteristic (ROC) analysis were performed. Results Among 150 patients, en bloc dissection within the submucosal tunnel was achieved in 100% of cases; en bloc retrieval through the lumen was achieved in 128 patients (85.33%), while 22 (14.67%) required piecemeal retrieval. Longitudinal length was the only independent predictor of piecemeal retrieval (aOR: 1.48 per 10 mm, 95% CI: 1.04-2.11, p = 0.031) and procedural difficulty (aOR = 1.07 per mm, 95% CI: 1.02-1.12, p = 0.003). Deep muscularis propria origin was the independent predictor of complication (aOR = 114.45, 95% CI: 30.13-434.71, p < 0.001). ROC analysis established optimal cut-offs: 43 mm for piecemeal retrieval (area under the curve [AUC] = 0.986), 38 mm for difficult procedures (AUC = 0.770), and 34 mm for complications (AUC = 0.739). Conclusions STER demonstrates high efficacy and safety for treating muscularis propria-originating SELs. Longitudinal length is the strongest predictor of piecemeal retrieval and procedural difficulty, while a deep-layer origin predicts complications. An internally validated, freely accessible interactive risk calculator ( https://nomosmed.streamlit.app/ ) is provided as a research prototype to support preoperative risk stratification; external multicenter validation is required before routine clinical adoption.